Optimising traffic flow in urban Accra through machine learning-based traffic light management

Abstract

Traffic congestion is a phenomenon that happens around the world, including urban Accra. Despite the construction of new roads, not much has been done to reduce traffic congestion in the city. This paper presents the development of an Intelligent Traffic Management System with the aim of optimising traffic flow and reducing traffic congestion in Accra. The system consists of a web interface to control traffic, a simulation environment to model traffic in urban Accra, and a machine learning model to predict when to switch to the optimal traffic signal. The machine learning model, based on Deep Q-Networks (DQN), predicts when to switch traffic phases in a given situation. The simulation environment was implemented using SUMO, modelling real-world traffic scenarios in selected areas of Accra to test and validate the system. The web interface provides a way to interact with the simulation for traffic control.